Tune a bit the fitting method (I'm still not completely convinced by its robustness, ideas are welcome)
Update the 4FGL catalog revision to 22.
Some changes for DESY's cluster.
Discovered feature: if you pass the 4FGL to the ScienceTools as a filter, then it takes its GTI intervals and you basically get a 4FGL-like analysis. This could be in principle done also with the old GTI-bin analysis then consolidate, but with that method it takes ages and it is very analysis crash-prone. I provide a first implementation to allow to use this.